Cluster returns outdated CLUSTER SLOTS

[andydunstall]

When Dragonfly finishes a migration, the source and target nodes are updated with the updated cluster config

This means the source can send a MOVED redirect to the target node for the migrated slots

However CLUSTER SLOTS still returns the old configuration, meaning if a client receives a MOVED redirect from shard A to shard B, then looks up CLUSTER SLOTS, CLUSTER SLOTS returns shard A still owns the slot

(We discussed in more depth in chat so won't copy here...)

[adiholden]

I am closing this issue, as discussed on vc. The change should be in cluster manager to pool more frequently the migration status and update the nodes with the new config accordingly.
Untill the update config will be sent to nodes the server will return with the relevant moved error which the client should follow up on to redirect the request